New benchmark LDU-Bench evaluates multimodal LLMs for lithography defect analysis

A new benchmark called LDU-Bench has been developed to evaluate multimodal large language models (MLLMs) in the context of lithography defect understanding. This benchmark, constructed from real industrial images, breaks down the review process into four tasks: defect triage, morphology recognition, coarse localization, and image-conditioned cause analysis. Current MLLMs show promise in defect triage but struggle with downstream tasks like morphology alignment and evidence-to-cause mapping, indicating a need for improved structured understanding. AI

IMPACT This benchmark aims to improve the usability and identify failure points of industrial MLLMs in critical lithography review processes.

RANK_REASON The item describes a new benchmark for evaluating multimodal LLMs in a specific industrial domain, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• anomaly detection
• integrated-circuit
• LDU-Bench
• lithography
• multimodal large language models

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →